Content uploaded by Jiacun Wang
Author content
All content in this area was uploaded by Jiacun Wang on May 24, 2021
Content may be subject to copyright.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 33
k
k k
k
33
3
Real-Time Operating Systems
e heart of many computerized embedded systems is real-time operating
system (RTOS). An RTOS is an operating system that supports the con-
struction of applications that must meet real-time constraints in addition to
providing logically correct computation results. It provides mechanisms and
services to carry out real-time task scheduling, resource management, and
intertask communication. In this chapter, we briefly review the main functions
of general-purpose operating systems, and then we discuss the characteristics
of RTOS kernels. After that, we introduce some widely used RTOS products.
3.1 Main Functions of General-Purpose Operating
Systems
An operating system (OS) is the software that sits between the hardware of
a computer and software applications running on the computer. An OS is a
resource allocator and manager. It manages the computer hardware resources
and hides the details of how the hardware operates to make the computer
system more convenient to use. e main hardware resources in a computer
are processor, memory, I/O controllers, disks, and other devices such as
terminals and networks.
An OS is a policy enforcer. It defines the rules of engagement between the
applications and resources and controls the execution of applications to prevent
errors and improper use of the computer.
An OS is composed of multiple software components, and the core com-
ponents in the OS form its kernel. e kernel provides the most basic level of
control over all of the computer’s hardware devices. e kernel of an OS always
runs in system mode, while other parts and all applications run in user mode.
Kernel functions are implemented with protection mechanisms such that they
could not be covertly changed through the actions of software running in user
space.
Real-Time Embedded Systems, First Edition. Jiacun Wang.
© 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 34
k
k k
k
34 3 Real-Time Operating Systems
e OS also provides an application programming interface (API), which
defines the rules and interfaces that enable applications to use OS features
and communicate with the hardware and other software applications. User
processes can request kernel services through system call or by message pa ssing.
With the system call approach, the user process applies traps to the OS routine
that determines which function is to be invoked, switches the processor to
system mode, calls the function as a procedure, and then switches the processor
back to user mode when the function completes and returns control to the
user process. In the message passing approach, the user process constructs a
message that describes the desired service and then uses a send function to
pass it to an OS process. e send function checks the desired service specified
in the message, changes the processor mode to system mode, and delivers it to
the process that implements the service. Meanwhile, the user process waits for
the result of the service request with a message receive operation. When the
service is completed, the OS process sends a message back to the user process.
In the rest of this section, we briefly introduce some of the main functions of
a typical general-purpose OS.
3.1.1 Process Management
Aprocess is an instance of a program in execution. It is a unit of work within
the system. A program is a passive entity, while a process is an active entity.
A process needs resources, such as CPU, memory, I/O devices, and files, to
accomplish its task. Executing an application program involves the creation of
a process by the OS kernel that assigns memory space and other resources,
establishes a priority for the process in multitasking systems, loads program
binary code into memory, and initiates execution of the application program,
which then interacts with the user and with hardware devices. When a process
is terminated, any reusable resources are released and returned to the OS.
Starting a new process is a heavy job for the OS, which includes allocating
memory, creating data structures, and coping code.
A thread is a path of execution within a process and the basic unit to
which the OS allocates processor time. A process can be single-threaded or
multithreaded. eoretically, a thread can do anything a process can do. e
essential difference between a thread and a process is the work that each one
is used to accomplish. reads are used for small tasks, whereas processes
are used for more heavyweight tasks – basically the execution of applications.
erefore, a thread is often called a lightweight process.
reads within the same process share the same address space, whereas
different processes do not. reads also share global and static variables, file
descriptors, signal bookkeeping, code area, and heap. is allows threads to
read from and write to the same data structures and variables and also facil-
itates communication between threads. us, threads use far less resources
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 35
k
k k
k
3.1 Main Functions of General-Purpose Operating Systems 35
(a) (b)
Code
Registers Stack
FilesData
Thread
Code
Registers
Stack
FilesData
Registers Registers
Stack Stack
Figure 3.1 (a) Single-threaded process; (b) Multithreaded process.
compared to processes. However, each thread of the same process has its own
thread status, program counter, registers, and stack, as illustrated in Figure 3.1.
A thread is the smallest unit of work managed independently by the scheduler
of the OS. In RTOSs, the term tasks is often used for threads or single-threaded
processes. For example, VxWorks and MicroC/OS-III are RTOSs that use the
term tasks . In this book, processes (threads) and tasks are used interchangeably.
Processes may create other processes through appropriate system calls, such
as fork or spawn. e process that does the creating is called the parent of the
other process, which is called its child. Each process is assigned a unique integer
identifier, call process identifier, or PID for short, when it is created. A process
can be created with or without arguments. Generally, a parent process is in
control of a child process. e parent can temporarily stop the child, cause it
to be terminated, send it messages, look inside its memory, and so on. A child
process may receive some amount of shared resources from its parent.
Processes may request their own termination by making the exit() system
call. Processes may also be terminated by the system for a variety of reasons,
such as the inability of the system to deliver necessary system resources, or
in response to a kill command or other unhandled process interrupt. When a
process terminates, all of its system resources are freed up, and open files are
flushed and closed.
Aprocesscanbesuspended for a variety of reasons. It can be swapping –the
OS needs to release sufficient main memory to bring in a process that is ready
to execute, or timing – a process that is executed periodically may be suspended
while waiting for the next time interval. A parent process may also wish to
suspend the execution of a child to examine or modify it or to coordinate the
activity of various children.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 36
k
k k
k
36 3 Real-Time Operating Systems
Sometimes, processes need to communicate with each other while they are
running. is is called interprocess communication (IPC). e OS provides
mechanisms to support IPC. Files, sockets, message queues, pipes, named
pipes, semaphores, shared memory, and message passing are typical IPC
mechanisms.
3.1.2 Memory Management
Main memory is the most critical resource in a computer system in terms of
speed at which programs run. e kernel of an OS is responsible for all system
memory that is currently in use by programs. Entities in memory are data and
instructions.
Each memory location has a physical address. In most computer architec-
tures, memory is byte-addressable, meaning that data can be accessed 8 bits at a
time, irrespective of the width of the data and address buses. Memory addresses
are fixed-length sequences of digits. In general, only system software such as
Basic Input/Output System (BIOS) and OS can address physical memory.
Most application programs do not have knowledge of physical addresses.
Instead, they use logic addresses. A logical address is the address at which a
memory location appears to reside from the perspective of an executing appli-
cation program. A logical address may be different from the physical address
due to the operation of an address translator or mapping function.
In a computer supporting virtual memory, the term physical address is
used mostly to differentiate from a virtual address. In particular, in computers
utilizing a memory management unit (MMU) to translate memory addresses,
the virtual and physical addresses refer to an address before and after
translation performed by the MMU, respectively. ere are several reasons
to use virtual memory. Among them is memory protection. If two or more
processes are running at the same time and use direct addresses, a memory
error in one process (e.g., reading a bad pointer) could destroy the memory
being used by the other process, taking down multiple programs due to a
single crash. e virtual memory technique, on the other hand, can ensure
that each process is running in its own dedicated address space.
Before a program is loaded into memory by the loader,partoftheOS,it
must be converted into a load module andstoredondisk.Tocreateaload
module, the source code is compiled by the compiler. e compiler produces
an object module. An object module contains a header that records the size
of each of the sections that follow, a machine code section that contains the
executable instructions compiled by the compiler, an initialized data section
that contains all data used by the program that require initialization, and the
symbol table section that contains all external symbols used in the program.
Some external symbols are defined in this object module and will be referred
to by other object modules, and some are used in this object module and
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 37
k
k k
k
3.1 Main Functions of General-Purpose Operating Systems 37
Figure 3.2 Structure of object module.
Header information
Machine code
Initialized data
Symbol table
Object module
Relocation info
defined in other object modules. e relocation information is used by the
linker to combine several object modules into a load module. Figure 3.2 shows
the structure of object modules.
When a process is started, the OS allocates memory to the process and then
loads the load module from disk into the memory allocated to the process.
During the loading, the executable code and initialized data are copied into the
process’ memory from the load module. In addition, memory is also allocated
for uninitialized data and runtime stack that is used to keep information about
each procedure call. e loader has a default initial size for the stack. When
the stack is filled up at runtime, extra space will be allocated to it, as long as
the predefined maximum size is not exceeded.
Many programming languages support memory allocation while a program
is running. It is done by the call of new in C++ and Java or malloc in C, for
example. is memory comes from a large pool of memory called the heap or
free store. At any given time, some parts of the heap are in use, while some are
free and thus available for future allocation. Figure 3.3 illustrates the memory
areas of a running process, in which the heap area is the memory allocated by
the process at runtime.
To avoid loading big executable files into memory, modern OS provides two
services: dynamic loading and dynamic linking. In dynamic loading, a routine
(library or other binary module) of a program is not loaded until it is called by
theprogram.Allroutinesarekeptondiskinarelocatableloadformat.emain
program is loaded into memory and is executed. Other routine methods or
modules are loaded on request. Dynamic loading makes better memory space
utilization, and unused routines are never loaded. Dynamic loading is useful
when a large amount of code is needed to handle infrequently occurring cases.
In dynamic linking, libraries are linked at execution time. is is compared
to static linking, in which libraries are linked at compile time, and thus,
the resultant executable code is big. Dynamic linking refers to resolving
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 38
k
k k
k
38 3 Real-Time Operating Systems
Process memory
Executable code
Initialized data
Uninitialized data
Heap
Stack
Figure 3.3 Memory areas of process.
symbols – associating their names with addresses or offsets – after compile
time. It is particularly useful for libraries.
e simplest memory management technique is single contiguous allocation.
at is, except the area reserved for the OS, all the memory is made available
toasingleapplication.isisthetechniqueusedintheMS-DOSsystem.
Another technique is partitioned allocation. It divides memory into multiple
memory partitions; each partition can have only one process. e memory
manager selects a free partition and assigns it to a process when it starts and
deallocates it when it is finished. Some systems allow a partition to be swapped
out to secondary storage to free additional memory. It is brought back into
memory for continued execution later.
Paged allocation divides memory into fixed-size units called page frames and
the program’s virtual address space into pages of the same size. e hardware
MMU maps pages to frames. e physical memory can be allocated on a
page basis while the address space appears to be contiguous. Paging does not
distinguish and protect programs and data separately.
Segmented memory allows programs and data to be broken up into logically
independent address spaces and to aid sharing and protection. Segments are
areas of memory that usually correspond to a logical grouping of information
such as a code procedure or a data array. Segments require hardware support
in the form of a segment table, which usually contains the physical address of
the segment in memory, its size, and other data such as access protection bits
and status (swapped in, swapped out, etc.).
As processes are loaded and removed from memory, the long, contiguous
free memory space becomes fragmented into smaller and smaller contiguous
pieces. Eventually, it may become impossible for the program to obtain large
contiguous chunks of memory. e problem is called fragmentation. In general,
smaller page size reduces fragmentation. e negative side is that it increases
the page table size.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 39
k
k k
k
3.1 Main Functions of General-Purpose Operating Systems 39
3.1.3 Interrupts Management
An interrupt is a signal from a device attached to a computer or from a running
process within the computer, indicating an event that needs immediate
attention. e processor responds by suspending its current activity, saving
its state, and executing a function called an interrupt handler (also called
interrupt service routine, ISR) to deal with the event.
Modern OSs are interrupt-driven. Virtually, all activities are initiated by
the arrival of interrupts. Interrupt transfers control to the ISR, through the
interrupt vector, which contains the addresses of all the service routines.
Interrupt architecture must save the address of the interrupted instruction.
Incoming interrupts are disabled while another interrupt is being processed.
A system call is a software-generated interrupt caused either by an error or by
auserrequest.
3.1.4 Multitasking
Real-world events may occur simultaneously. Multitasking refers to the
capability of an OS that supports multiple independent programs running on
the same computer. It is mainly achieved through time-sharing, which means
that each program uses a share of the computer’s time to execute. How to
share processors’ time among multiple tasks is addressed by schedulers,which
follow scheduling algorithms to decide when to execute which task on which
processor.
Each task has a context, which is the data indicating its execution state and
stored in the task control block (TCB), a data structure that contains all the
information that is pertinent to the execution of the task. When a scheduler
switches a task out of the CPU, its context has to be stored; when the task
is selected to run again, the task’s context is restored so that the task can be
executed from the last interrupted point. e process of storing and restoring
the context of a task during a task switch is called a context switch,whichis
illustrated in Figure 3.4.
Context switches are the overhead of multitasking. ey are usually com-
putationally intensive. Context switch optimization is one of the tasks of OS
design. is is particularly the case in RTOS design.
3.1.5 File System Management
Files are the fundamental abstraction of secondary storage devices. Each file is
anamedcollectionofdatastoredinadevice.Animportantcomponentofan
OS is the file system, which provides capabilities of file management, auxiliary
storage management, file access control, and integrity assurance.
File management is concerned with providing the mechanisms for files to be
stored, referenced, shared, and secured. When a file is created, the file system
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 40
k
k k
k
40 3 Real-Time Operating Systems
Task A Task B
OS
Interrupt or system call
Save state into TCBa
Restore state from TCBb
Save state into TCBb
Restore state from TCBa
Executing
Idle
Executing
Executing
Idle
Idle
Figure 3.4 Context switch between tasks A and B.
File Location
File directory
Data Data
Null
Data
Data
Data Data
Null
Data
Figure 3.5 The block chaining scheme in file storage.
allocates an initial space for the data. Subsequent incremental allocations follow
as the file grows. When a file is deleted or its size is shrunk, the space that is
freed up is considered available for use by other files. is creates alternating
usedandunusedareasofvarioussizes.Whenafileiscreatedandthereisnot
an area of contiguous space available for its initial allocation, the space must
be assigned in fragments. Because files do tend to grow or shrink over time,
and because users rarely know in advance how large their files will be, it makes
sense to adopt noncontiguous storage allocation schemes. Figure 3.5 illustrates
the block chaining scheme. e initial address of storage of a file is identified
by its file name.
Typically, files on a computer are organized into directories, which constitute
a hierarchical system of tree structure.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 41
k
k k
k
3.1 Main Functions of General-Purpose Operating Systems 41
A file system typically stores necessary bookkeeping information for each file,
including the size of data contained in the file, the time the file was last modified,
its owner user ID and group ID, and its access permissions.
e file system also provides a spectrum of commands to read and write the
contents of a file, to set the file read/write position, to set and use the protection
mechanism, to change the ownership, to list files in a directory, and to remove
a file.
File access control can be realized using a two-dimensional matrix that lists
all users and all files in the system. e entry at index (i,j) of the matrix specifies
if user iis allowed to access file j.Inasystemthathasalargenumberofusers
and contains a large number of files, this matrix would be very large and very
sparse.
A scheme that requires much less space is to control access to various
user classes. Role-based access control (RBAC) is an access control method
where access to data is performed by authorized users. RBAC assigns users
to specific roles, and permissions are granted to each role based on the user’s
job requirements. Users can be assigned a number of roles in order to conduct
day-to-day tasks. For example, a user may need to have a developer role as well
as an analyst role. Each role would define the permissions that are needed to
access different objects.
3.1.6 I/O Management
Modern computers interact with a wide range of I/O devices. Keyboards, mice,
printers, disk drives, USB drives, monitors, networking adapters, and audio
systemsareamongthemostcommonones.OnepurposeofanOSistohide
peculiarities of hardware I/O devices from the user.
In memory-mapped I/O, each I/O device occupies some locations in the
I/O address space. Communication between the I/O device and the processor
is enabled through physical memory locations in the I/O address space. By
reading from or writing to those addresses, the processor gets information
from or sends commands to I/O devices.
Most systems use device controllers. A device controller is primarily an
interface unit. e OS communicates with the I/O device through the device
controller. Nearly all device controllers have direct memory access (DMA)
capability, meaning that they can directly access the memory in the system,
without the intervention by the processor. is frees up the processor of the
burden of data transfer from and to I/O devices.
Interrupts allow devices to notify the processor when they have data to
transfer or when an operation is complete, allowing the processor to perform
other duties when no I/O transfers need its immediate attention. e processor
has the interrupt request line that it senses after executing every instruction.
When a device controller raises an interrupt by asserting a signal on the
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 42
k
k k
k
42 3 Real-Time Operating Systems
interrupt request line, the processor catches it and saves the state and then
transfers control to the interrupt handler. e interrupt handler determines
the cause of the interrupt, performs necessary processing, and executes a
return from interrupt instruction to return control to the processor.
I/O operations often have high latencies. Most of this latency is due to the
slow speed of peripheral devices. For example, information cannot be read from
or written to a hard disk until the spinning of the disk brings the target sectors
directly under the read/write head. e latency can be alleviated by having one
or more input and output buffers associated with each device.
3.2 Characteristics of RTOS Kernels
Although a general-purpose OS provides a rich set of services that are also
needed by real-time systems, it takes too much space and contains too many
functions that may not be necessary for a specific real-time application.
Moreover, it is not configurable, and its inherent timing uncertainty offers no
guarantee to system response time. erefore, a general-purpose OS is not
suitable for real-time embedded systems.
ere are three key requirements of RTOS design. Firstly, the timing behavior
of the OS must be predictable. For all services provided by the OS, the upper
bound on the execution time must be known. Some of these services include
OS calls and interrupt handling. Secondly, the OS must manage the timing and
scheduling, and the scheduler must be aware of task deadlines. irdly, the OS
must be fast. For example, the overhead of context switch should be short. A
fast RTOS helps take care of soft real-time constraints of a system as well as
guarantees hard deadlines.
As illustrated in Figure 3.6, an RTOS generally contains a real-time kernel
and other higher-level services such as file management, protocol stacks,
a Graphical User Interface (GUI), and other components. Most additional
services revolve around I/O devices. A real-time kernel is software that
manages the time and resources of a microprocessor or microcontroller and
provides indispensable services such as task scheduling and interrupt handling
to applications. Figure 3.7 shows a general structure of a microkernel. In
embedded systems, a small amount of code called board support package
(BSP) is implemented for a given board that conforms to a given OS. It is
commonly built with a bootloader that contains the minimal device support to
load the OS and device drivers for all the devices on the board.
In the rest of this section, we introduce some most important real-time
services that are specified in POSIX 1.b for RTOS kernels.
3.2.1 Clocks and Timers
Most embedded systems must keep track of the passage of time. e length
of time is represented by the number of system ticks in most RTOS kernels.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 43
k
k k
k
3.2 Characteristics of RTOS Kernels 43
RTOS
Applications
Kernel
Timing scheduling
Interrupts handling
Memory mgmt
File
management
GUI
Device I/O
Protocol stacks
Debugging
tools
Other services
Board support package
Target hardware
Figure 3.6 A high-level view of RTOS.
External
interrupts
System
calls
Hardware/software
exceptions
Clock
interrupts
Immediate
interrupt
service
Case of
Create_thread
Suspend_thread
Destroy_thread
Create_timer
Timer_sleep
Timer_notify
Other system calls
.
.
.
.
.
.
Scheduling
Exception
handling
Time
service
Kernel
Figure 3.7 Structure of a microkernel.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 44
k
k k
k
44 3 Real-Time Operating Systems
e RTOS works by setting up a hardware timer to interrupt periodically, say,
every millisecond, and bases all timings on the interrupts. For example, if in a
task you call the taskDelay function in VxWorks with a parameter of 20, then
the task will block until the timer interrupts for 20 times. In the POSIX stan-
dard, each tick is equal to 10 milliseconds, and in each second, there are 100
ticks. Some RTOS kernels, such as VxWorks, define routines that allow the user
to set and get the value of system tick. e timer is often called a heartbeat
timer, and the interrupt is also called clock interrupt.
At each clock interrupt, an ISR increments tick count, checks to see if it is
now time to unblock or wake up a task. If this is the case, it calls the scheduler
to do the scheduling again.
Based on the system tick, an RTOS kernel allows you to call functions of your
choice after a given number of system ticks, such as the taskDelay function
in VxWorks. Depending upon the RTOS, your function may be directly called
from the timer ISR. ere are also other timing services. For example, most
RTOS kernels allow developers to limit how long a task will wait for a message
from a queue or a mailbox and how long a task will wait for a semaphore, and
so on.
Timers improve the determinism of real-time applications. Timers allow
applications to set up events at predefined intervals or time. POSIX specified
several timer-related routines, including the following:
•timer_create() – allocate a timer using the specified clock for a timing
base.
•timer_delete() – remove a previously created timer.
•timer_gettime() – get the remaining time before expiration and the
reload value.
•timer_getoverrun() – return the timer expiration overrun.
•timer_settime() – set the time until the next expiration and arm timer.
•nanosleep() – suspend the current task until the time interval elapses.
3.2.2 Priority Scheduling
Because real-time tasks have deadlines, being soft or hard, all tasks are not
equal in terms of the urgency of getting executed. Tasks with shorter deadlines
should be scheduled for execution sooner than those with longer deadlines.
erefore, tasks are typically prioritized in an RTOS. Moreover, if a higher
priority task is released while the processor is serving a lower priority task,
the RTOS should temporarily suspend the lower priority task and immediately
schedule the higher priority on the processor for execution, to ensure that the
higher priority task is executed before its deadline. is process is called pre-
emption. Task scheduling for real-time applications is typically priority-based
and preemptive. Examples are earliest deadline first (EDF) scheduling and
rate monotonic (RM) scheduling. Scheduling algorithms that do not take task
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 45
k
k k
k
3.2 Characteristics of RTOS Kernels 45
Ready tasks
Scheduler Dispatcher
Task control blocks
Context switcher CPU
Remove the running task
Figure 3.8 Priority scheduling.
priorities into account, such as first-in-first-service and round-robin, are not
suitable for real-time systems.
In priority-driven preemptive scheduling, the preemptive scheduler has a
clock interrupt task that provides the scheduler with options to switch after
the task has had a given period to execute – the time slice. is scheduling
system has the advantage of making sure that no task hogs the processor for
any time longer than the time slice.
As shown in Figure 3.8, an important component involved in scheduling is
the dispatcher, which is the module that gives control of the CPU to the task
selectedbythescheduler.Itreceivescontrolinkernelmodeastheresultof
an interrupt or system call. It is responsible for performing context switches.
edispatchershouldbeasfastaspossible,sinceitisinvokedduringeverytask
switch. During the context switches, the processor is virtually idle for a fraction
of time; thus, unnecessary context switches should be avoided.
e key to the performance of priority-driven scheduling is in choosing
priorities for tasks. Priority-driven scheduling may cause low-priority tasks to
starve and miss their deadlines. In the next two chapters, several well-known
scheduling algorithms and resource access control protocols will be discussed.
3.2.3 Intertask Communication and Resource Sharing
In an RTOS, a task cannot call another task. Instead, tasks exchange infor-
mation through message passing or memory sharing and coordinate their
execution and access to shared data using real-time signals, mutex, or
semaphore objects.
3.2.3.1 Real-Time Signals
Signals are similar to software interrupts. In an RTOS, a signal is automatically
sent to the parent when a child process terminates. Signals are also used for
many other synchronous and asynchronous notifications, such as waking a
process when a wait call is performed and informing a process that it has
issued a memory violation.
POSIX extended the signal generation and delivery to improve the real-time
capabilities. Signals take an important role in real-time systems as the
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 46
k
k k
k
46 3 Real-Time Operating Systems
way to inform the processes of the occurrence of asynchronous events
such as high-resolution timer expiration, fast interprocess message arrival,
asynchronous I/O completion, and explicit signal delivery.
3.2.3.2 Semaphores
Semaphores are counters used for controlling access to resources shared
among processes or threads. e value of a semaphore is the number of units
of the resource that are currently available. ere are two basic operations
on semaphores. One is to atomically increment the counter. e other is to
wait until the counter is nonnull and atomically decrement it. A semaphore
tracks only how many resources are free; it does not keep track of which of the
resources are free.
A binary semaphore acts similarly to a mutex,whichisusedinthecasewhen
aresourcecanonlybeusedbyatmostonetaskatanytime.
3.2.3.3 Message Passing
In addition to signals and semaphores, tasks can share data by sending messages
in an organized message passing scheme. Message passing is much more useful
for information transfer. It can also be used just for synchronization. Message
passing often coexists with shared memory communication. Message contents
can be anything that is mutually comprehensible between the two parties in
communication. Two basic operations are send and receive.
Message passing can be direct or indirect. In direct message passing, each
process wanting to communicate must explicitly name the recipient or sender
of the communication. In indirect message passing, messages are sent to and
received from mailboxes or ports. Two processes can communicate in this way
only if they have a shared mailbox.
Message passing can also be synchronous or asynchronous.Insynchronous
message passing, the sender process is blocked until the message primitive
has been performed. In asynchronous message passing, the sender process
immediately gets control back.
3.2.3.4 Shared Memory
Shared memory is a method that an RTOS uses to map common physical space
into independent process-specific virtual space. Shared memory is commonly
used to share information (resources) between different processes or threads.
Shared memory must be accessed exclusively. erefore, it is necessary to use
mutex or semaphores to protect the memory area. e code segment in a task
that accesses the shared data is called a critical section. Figure 3.9 shows that
two tasks share a memory region.
A side effect in using shared memory is that it may cause priority inversion,a
situation that a low-priority task is running while a high-priority task is waiting.
More details of priority inversion will be discussed in Chapter 5.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 47
k
k k
k
3.2 Characteristics of RTOS Kernels 47
Task A Task B
.
.
.
wait(S)
critical
section
signal(s)
.
.
.
Shared
memory
.
.
.
wait(S)
critical
section
signal(s)
.
.
.
Figure 3.9 Shared memory and critical section.
3.2.4 Asynchronous I/O
ere are two types of I/O synchronization: synchronous I/O and asyn-
chronous I/O. In synchronous I/O, when a user task requests the kernel for
I/O operation and the requested is granted, the system will wait until the
operation is completed before it can process other tasks. Synchronous I/O is
desirable when the I/O operation is fast. It is also easy to implement.
An RTOS supports the overlap of application processing and application-
initiated I/O operations. is is the RTOS service of asynchronous I/O. In
asynchronous I/O, after a task requests I/O operation, while this task is waiting
for I/O to complete, other tasks that do not depend on the I/O results will
be scheduled for execution. Meanwhile, tasks that depend on the I/O having
completed are blocked. Asynchronous I/O is used to improve throughput,
latency, and/or responsiveness.
Figure 3.10 illustrates the idea of synchronous I/O and asynchronous I/O.
3.2.5 Memory Locking
Memory locking is a real-time capability specified by POSIX that is intended
for a process to avoid the latency of fetching a page of memory. It is achieved
by locking the memory so that the page is memory-resident, that is, it remains
in the main memory. is allows fine-grained control of which part of the
application must stay in physical memory to reduce the overhead associated
with transferring date between memory and disk. For example, memory
locking can be used to keep in memory a thread that monitors a critical
processthatrequiresimmediateattention.
When the process exits, the locked memory is automatically unlocked.
Locked memory can also be unlocked explicitly. POSIX, for example, defined
mlock() and munlock() functions to lock and unlock memory. e
munlock function unlocks the specified address range regardless of the
number of times the mlock function was called. In other words, you can lock
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 48
k
k k
k
48 3 Real-Time Operating Systems
Kernel mode
Synchronous I/O
Executing
a task
Idle
Executing Executing
User mode
I/O starts
I/O ends
Requests I/O
Task resumes Task resumes
Asynchronous I/O
Executing
a task
Executing
another task
User mode
I/O starts
I/O ends
Requests I/O
Interrupted
Kernel mode
Figure 3.10 Synchronous I/O versus asynchronous I/O.
address ranges over multiple calls to the mlock function, but the locks can be
removed with a single call to the munlock function. In other words, memory
locks don’t stack.
More than one process can lock the same or overlapping region, and in that
case, the memory region remains locked until all the processes have unlocked it .
3.3 RTOS Examples
3.3.1 LynxOS
e LynxOS RTOS is a Unix-like RTOS from Lynx Software Technologies.
LynxOS is a deterministic, hard RTOS that provides POSIX-conformant
APIs in a small-footprint embedded kernel. It features predictable worst-case
response time, preemptive scheduling, real-time priorities, ROMable kernel,
and memory locking. LynxOS provides symmetric multiprocessing support to
fully take advantage of multicore/multithreaded processors. LynxOS 7.0, the
latest version, includes new tool chains, debuggers, and cross-development
host support.
e LynxOS RTOS is designed from the ground up for conformance
to open-system interfaces. It leverages existing Linux, UNIX, and POSIX
programming talent for embedded real-time projects. Real-time system
development time is saved, and programmers are able to be more productive
using familiar methodologies as opposed to learning proprietary methods.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 49
k
k k
k
3.3 RTOS Examples 49
LynxOS is mostly used in real-time embedded systems, in applications for
avionics, aerospace, the military, industrial process control, and telecommuni-
cations. LynxOS has already been used in millions of devices.
3.3.2 OSE
OSE is an acronym for the operating system embedded. It is a real-time
embedded OS created by the Swedish information technology company ENEA
AB. OSE uses signals in the form of messages passed to and from processes
in the system. Messages are stored in a queue attached to each process. A link
handler mechanism allows signals to be passed between processes on separate
machines, over a variety of transports. e OSE signaling mechanism formed
the basis of an open-source interprocess kernel design.
e Enea RTOS family shares a high-level programming model and an
intuitive API to simplify programming. It consists of two products, each
optimized for a specific class of applications:
•Enea OSE is a robust and high-performance RTOS optimized for multicore,
distributed, and fault-tolerant systems.
•Enea OSEck is a compact and multicore DSP-optimized version of ENEA’s
full-featured OSE RTOS.
OSEsupportsmanymainly32-bitprocessors,suchasthoseinARM,
PowerPC, and MIPS families.
3.3.3 QNX
eQNXNeutrinoRTOSisafull-featuredandrobustOSisdevelopedby
QNX Software Systems Limited, a subsidiary of BlackBerry. QNX products are
designed for embedded systems running on various platforms, including ARM
and x86, and a host of boards implemented in virtually every type of embedded
environment.
As a microkernel-based OS, QNX is based on the idea of running most of
theOSkernelintheformofanumberofsmalltasks,knownasservers.is
differs from the more traditional monolithic kernel, in which the OS kernel is
a single, very large program composed of a huge number of components with
special abilities. In the case of QNX, the use of a microkernel allows developers
to turn off any functionality they do not require without having to change the
OS itself; instead, those servers will simply not run.
e BlackBerry PlayBook tablet computer designed by BlackBerry uses a
versionofQNXastheprimaryOS.eBlackBerrylineofdevicesrunningthe
BlackBerry 10 OS are also based on QNX.
3.3.4 VxWorks
VxWorks is an RTOS developed as proprietary software by Wind River, a
subsidiary company of Intel providing embedded system software, which
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 50
k
k k
k
50 3 Real-Time Operating Systems
comprises runtime software, industry-specific software solutions, simulation
technology, development tools, and middleware. As other RTOS products,
VxWorks is designed for use in embedded systems requiring real-time and
deterministic performance.
VxWorks supports Intel, MIPS, PowerPC, SH-4, and ARM architectures. e
VxWorks Core Platform consists of a set of runtime components and devel-
opment tools. VxWorks core development tools are compilers such as Diab,
GNU, and Intel C++ Compiler (ICC) and its build and configuration tools. e
system also includes productivity tools such as its Workbench development
suite and Intel tools and development support tools for asset tracking and host
support. Cross-compiling is used with VxWorks. Development is achieved on a
“host” system where an integrated development environment (IDE), including
the editor, compiler toolchain, debugger, and emulator, can be used. Software
is then compiled to run on the “target” system. is allows the developer to
work with powerful development tools while targeting more limited hardware.
VxWorks is used by products over a wide range of market areas: aerospace
and defenses, automotive, industrial such as robots, consumer electronics,
medical area, and networking. Several notable products also use VxWorks as
the onboard OS. Examples in the area of spacecraft are the Mars Reconnais-
sance Orbiter, the Phoenix Mars Lander, the Deep Impact Space Probe, and
the Mars Pathfinder.
3.3.5 Windows Embedded Compact
Formerly known as Windows CE, Windows Embedded Compact is an OS
subfamily developed by Microsoft as part of its Windows Embedded family
of products. It is a small-footprint RTOS and optimized for devices that have
minimal memory, such as industrial controllers, communication hubs, and
consumer electronics devices such as digital cameras, GPS systems, and also
automotive infotainment systems. It supports x86, SH (automotive only),
and ARM.
Exercises
1What is the kernel of an operating system? In which mode does it run?
2What are the two approaches that a user process interacts with the
operating system? Discuss the merits of each approach.
3What is the difference between a program and a process? What is the
difference between a process and a thread?
4Should terminating a process also terminate all its children? Give an
example when this is a good idea and another example when this is a
bad idea.
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 51
k
k k
k
3.3 RTOS Examples 51
5Should the open files of a process be automatically closed when the
process exits?
6What is the difference between an object module and a load module?
7Discuss the merits of each memory allocation techniques introduced in
this chapter.
8What do we mean when we say an OS is interrupt-driven? What does the
processor do when an interrupt occurs?
9Whatisacontextswitchandwhendoesitoccur?
10 Does a file have to be stored in a contiguous storage region in a disk?
11 What are the benefits of using memory-mapped I/O?
12 Why can a general-purpose OS not meet the requirements of real-time
systems?
13 How does memory fragmentation form? Name one approach to control it.
14 What are the basic functions of an RTOS kernel?
15 HowdoesanRTOSkeeptrackofthepassageoftime?
16 Why is it necessary to use priority-based scheduling in a real-time appli-
cation?
17 What are the general approaches that different tasks communicate with
each other and perform synchronization of their actions in access to
shared resources?
18 Compare synchronous I/O and asynchronous I/O and list their advan-
tages and disadvantages.
19 How does the memory locking technique improve the performance of a
real-time system?
Trim Size: 152mm x 229mm Single Column Wang c03.tex V1 - 05/13/2017 12:16pm Page 52
k
k k
k
52 3 Real-Time Operating Systems
Suggestions for Reading
Many textbooks are available that provide thorough and in-depth discussion
on general-purpose OSs. Examples are [1–3]. Cooling [4] provides a great
overview of the fundamentals of RTOS for embedded programming without
being specific to any one vendor. POSIX-specified real-time facilities are
described in Ref. [5]. More information regarding the commercial RTOS
products can be found on their official websites.
References
1Doeppner, T. (2011) Operating Systems in Depth,Wiley.
2McHoes, A.M. and Flynn, I.M. (2011) Understand Operating Systems,
6th edn, Course Technology Cengage Learning.
3Stallings, W. (2014) Operating Systems:Internals and Design Principles,
8th edn, Pearson.
4Cooling, J. (2013) Real-Time Operating Systems, Kindle edn, Lindentree
Associates.
5Gallmeister, B. (1995) POSIX 4.0:Programming for the Real World, O’Reilly
& Associates, Inc..
URLs
LynxOS, http://www.lynxos.com/
VxWorks, http://www.windriver.com/products/vxworks/
Windows CE, https://www.microsoft.com/windowsembedded/en-us/
windows-embedded-compact-7.aspx
QNX, http://www.qnx.com/content/qnx/en.html
OSE, http://www.enea.com/ose